Refactor GPU node spec in radixconfig

[nilsgstrabo]

We should simplify the node spec (https://www.radix.equinor.com/references/reference-radix-config/#node) in radixconfig.yaml.
Today the user sets gpu and gpuCount. radix-operator translates this into a toleration and a nodeAffinity selector.
The current node spec does not support definition VM types, e.g. with more/faster memory and cpus. It is bound to GPU types only.

I suggest we deprecate/delete the current node spec and replace with a "simple" string value, e.g. nodeType: nv-v100-2, where nv-v100-2 is in a fixed list of nodes types (mapped to a Azure VM size). We can then add new nodepools with new vm types and add the new vm type to the list of supported types.

Use labels to direct the pod to correct nodepool.

DoD:
In radixconfig it must be possible to add extra information to select different node types

I.e. bronze = VM type, Mem, CPU